consistent use of emulation across different stages …

Proceedings of the 2013 Winter Simulation Conference

R. Pasupathy, S.-H. Kim, A. Tolk, R. Hill, and M. E. Kuhl, eds.

CONSISTENT USE OF EMULATION ACROSS DIFFERENT STAGES OF PLANT

DEVELOPMENT – THE CASE OF DEADLOCK AVOIDANCE FOR CYCLIC CUT-TO-SIZE

PROCESSES

Ruth Fleisch

Robert Schoech

Thorsten Prante

Robert Pflegerl

V-Research GmbH Schelling Anlagenbau GmbH

Stadtstrasse 33 Gebhard-Schwärzler-Strasse 34

A-6850 Dornbirn, AUSTRIA A-6858 Schwarzach, AUSTRIA

ABSTRACT

This paper presents an emulation tool and the way in which it supports the test-driven design of zero-

deadlock cut-to-size plants. In contrast to previous plant layouts featuring a linear material flow, panels

can now be allocated to the same saw again for their further partitioning. The resulting circles in the mate-

rial flow of cut-to-size plants impose new challenges as they may cause deadlocks, which means that the

manufacturing system or parts of it remains indefinitely blocked. In order to prevent deadlock occurrenc-

es, an algorithm is developed and implemented into the software controlling the cut-to-size plant. The

necessary tests are performed in an iterative way with the help of our emulation tool, which connects the

plant control software with the virtual analogs of the control and the field layers.

1 INTRODUCTION

Cut-to-size plants are complex machinery, the purpose of which is to realize cutting, sorting, and stacking

of panels of different sizes and materials (e.g., wood, plastics, or metal). The just mentioned main cut-to-

size processes are carried out by cut-to-size saws, sorting carriages, stacking devices, roller tracks, etc., as

depicted in Figure 1.

Feeding Area

Cutting Area Sorting Area Stacking Area

Feeding Area

Cutting Area Sorting Area Stacking Area

Figure 1: Exemplary cut-to-size plant layout with units for feeding the plant, cutting, sorting, and stacking

grouped into areas

The illustrated plant components – more precisely, devices and units – can be combined and config-

ured in a variety of ways, leading to a high degree of structural complexity. This diversification by varia-

tion is essential for matching the manifold and specialized requirements of a whole range of customers

from industry sectors such as furniture manufacturing and circuit board industry. On the other hand, along

2565978-1-4799-2076-1/13/$31.00 ©2013 IEEE

Fleisch, Schoech, Prante, and Pflegerl

with the structural complexity comes the exposition of complex dynamic behavior by cut-to-size plants.

Examples include parallel material movements, buffer areas, a process-inherent increase of panels in such

plants, and a large number of simultaneous activities, in general.

In view of these complexities, the ultimate motivation of the here presented work is to incrementally

develop a model-based, multi-faceted decision support tool, which consistently uses and refines – across

stages of plant design and lifecycle – “an emulation model (as distinct from a pure simulation model)

(…), where some functional part of the model is carried out by a part of the real system” (McGregor

2002). To be cost effective, this tool should at best be applicable all the way from concept planning, via

bid preparation and sales negotiation, through design, testing, and deployment until operation and adjust-

ments. Concretely, in Schoech, Schmid, and Hillbrand (2011) and Schoech, Schmid, and Hillbrand (to

appear), it was already shown that the decision support tool provides added value for both suppliers and

clients during sales processes of cut-to-size plants. The particular motivation for that came from the fact

that typical customers in this highly specialized business – the discussed plants are built according to cus-

tomer specifications (lot-size one, make-to-order strategy) – commonly place an order for a turnkey plant

only after complex sales procedures, in which the plant supplier needs to demonstrate the capability to

fulfill specific customer requirements. Consequently, the following has been demonstrated: Support for

iterative requirements specification and plant conception, plus accurate calculation of critical performance

indicators, and 3D animation of material flow (amongst others reducing information asymmetries be-

tween manufacturer and potential customer).

On the other hand, in this paper, usefulness of the same model-based tool in later phases of design

and implementation will be shown. In particular, the focus will be on the support for trade-off decisions

such as ensuring absence of deadlocks while still fulfilling performance-indicator requirements for cut-to-

size plants, which feature feedback in their material flow. This is closely related to plant control software

implementation.

2 RELATED WORK

While the deployment of the here presented emulation approach during sales processes of cut-to-size

plants and the related benefits with regard to optimizing plant layout decisions have already been dis-

cussed in Schoech, Schmid, and Hillbrand (2011) and Schoech, Schmid, and Hillbrand (to appear), in this

paper the use of the emulation tool during a later stage of plant development is described, where it sup-

ports establishing a deadlock-handling strategy in the process control layer of a cut-to-size plant. A simi-

lar approach of virtually testing upper-level control software is also adopted by Okolnishnikov (2011),

who gives examples of developing process control systems with the use of emulation models in various

industrial fields such as a railway tunnel or underground coal mining. In Meyer, Pöge, and Mayer (2012)

the expansion of an existing simulation object library for the automotive industry to include emulation

functionality is described. In that project the simulated plant is linked to the manufacturing execution sys-

tem (MES) as well, whereas Seidel, Donath, and Haufe (2012) present an integrated simulation and virtu-

al commissioning environment for controls of (intralogistics) material handling systems which comprises

emulations not only connected to the upper level of a material flow controller (MFC) but also to the lower

level of programmable logic controllers (PLC).

The following section gives an explanation of the emulation approach used in our tool together with

the different layers of the control system. Section 4 shows how this conceptual model is implemented re-

sulting in a tool which enables a domain expert of cut-to-size plants to model, emulate, and animate a

plant and its processes. In section 5, we take a closer look at plant layouts with feedback in their material

flows, before we provide details about deploying our emulation tool to design zero-deadlock plants for

such layouts in section 6.

3 EMULATION APPROACH

According to Mertens (2001), the term emulation can be defined as the virtual reproduction of certain as-

pects of hardware or software systems on another system. Therefore, emulation can be seen as a special

2566


case of simulation, supplemented with the coupling of real-world functional components. For emulation

of cut-to-size plants, orders from the plant server are executed within a virtual plant.

As Figure 2 illustrates, classical plants are centrally and hierarchically controlled real time systems

(Günthner and ten Hompel 2010). Within such real-world systems the field layer, constituting the lowest

level, represents all mechanical components with their actuators and sensors and controls the material

handling. The control layer resides above the process level. On this layer, sensor data are processed and

control signals for the actuators are generated. This level represents the basis for an automated, modular

plant. Moreover, it coordinates the handover of load data and controls the material flow of the plant mod-

ules. The process control layer is the highest layer in the control pyramid and is often called plant server.

The control layer and the process control layer are connected by means of a communication channel,

which passes the scheduled orders to the control layer and receives confirmation when all actions have

been processed.

Process control layer

Field layer

Control layer Virtual control layer

Virtual field layer

Emulation

Reality

Simulation

Process control layer

Field layer

Control layer Virtual control layer

Virtual field layer

Emulation

Reality

Simulation

Figure 2: Central hierarchical control systems

The plant logic of the real-world system as described above is mapped to a model within the virtual

plant: The virtual field layer of the emulated system visualizes all mechanical components and displays

the kinematic movements of the material handling within a virtual environment. The virtual control layer

prepares all orders originating from the process control layer and controls each virtual plant unit. These

are one-to-one correspondents of real-world plant units. Since there is a tight coupling between real-world

and simulation systems, the virtual system can be called an emulation system (McGregor 2002).

In contrast to simulation, emulation models are mostly used in much more precisely defined ways, for

example, “in order to test the operation of the control system under different system loading conditions”

(McGregor 2002). The conditions under which the evaluations are carried out can be better controlled, al-

lowing the study of different scenarios the control system has to deal with (McGregor 2002).

4 FRAMEWORK: ARCHITECTURE AND USAGES

This section presents a general overview of the system architecture for our emulation-based decision sup-

port tool. The foundation of our system, the in-house developed software framework Application Plat-

form Simulation (Dobler, Saler, and Maerz 2008), provides simulation libraries with generic modeling

functions which can be used to implement domain-specific modeling environments. A number of domain-

specific modeling methods and applications, such as planning of transportation networks (Hillbrand and

Schmid 2011) or warehousing structures (Maerz and Saler 2008), have already been implemented on top

of this generic platform. For the work presented in this paper, specific methods for cut-to-size plants were

implemented. The main intention has been to enable domain experts (e.g., technical sales personnel) to

define and project alternative plant configurations within the virtual system and to evaluate them against

the predefined critical performance indicators in order to identify an optimal solution. By using a domain-

specific modeling environment with an underlying simulation model built-in, the domain expert can take

advantage of emulation techniques without needing to have simulation expertise. Thereto, the software

architecture of the decision support system for cut-to-size-plants is arranged in two levels (illustrated left

2567


and right of the dashed vertical line in the left part of Figure 3): the modeling and results level and the

emulation and visualization level.

Modeling & resultsModeling & results Emulation & visualizationEmulation & visualization

Reality

Emulation

Plant server

Simulator

Simulation modelScheduler

Templates

System model

Results

Middleware

SynchronizationLogic of

decomposing orders

Part lists with

cutting pattern,

sorting & stacking

data

Virtual field layer Kinematics

x

y

z

Virtual control layer Task sequencesLiftConvey

Process control layer OrdersDrive-In to

turning device

Turn object on

turning device

Order handling

Figure 3: System architecture (left side) and order handling (right side)

Firstly, the modeling and results level is used for the development of unit-based system models (Zei-

gler and Sarjoughian 2003): a plant is designed as a collection of plant units. A plant unit template repre-

sents a number of variations of a self-contained plant unit, which again consist of a set of devices. A de-

vice is specified by attributes which describe process-relevant and mechanical information, velocities and

meta-information for the instantiation of each class type used in the virtual field layer and in the virtual

control layer. These unit templates represent atomic modules of the virtual plant and cannot be further

broken down. They are linked together in drag-and-drop manner to form a system model of the plant in

question in a graphical editor (see Figure 5 in Schoech, Schmid, and Hillbrand (2011)). The work flow of

the plant is modeled implicitly by the coupling of the individual plant units. Once the plant is specified as

a unit-based system model, an emulation run and the visualization of the material flow are triggered at the

modeling and results level. For transforming the system model into a simulation model, the unit-based

system model is rendered persistent as a well-formed XML document. The transformation process of

building a simulation model is automated and consists of two parts, depicted by the two arrows crossing

the vertical dashed line in Figure 3 from left to right: creation of simulation objects and the assignment of

values to all public properties, on the one hand. Additionally, instantiation of the virtual control layer

takes place depending on the defined unit-based system model description (upper arrow).

The emulation and visualization level represents the plant server and the virtual system. Instead of

controlling the real plant, the plant server controls the virtual one. After starting the emulation, the plant

server broadcasts orders to the virtual control layer, which are generated on the basis of part lists with cut-

ting patterns as well as sorting and stacking patterns for the panels. On the virtual control layer, task se-

quences are generated from the orders (see Figure 3 on the right side). A single task describes an activity

that imitates a physical mechanism which is executed on a simulation object in the simulation model. The

virtual field layer, which is based on the simulation engine FlexSim©, retrieves the task sequences and

transforms them into discrete events. The simulation model implements 3D visualization of the devices of

the plant units and of the panels as well as of their kinematic flows according to the tasks. After execution

of the tasks of an order, the confirmation of the order is sent to the plant server via the virtual control lay-

er. Depending on the order confirmation, new orders are triggered for processing task sequences.

While running the emulation, run time information is logged, which afterwards can be viewed and

analyzed by the user in the modeling and results level.

2568


An important issue to be considered with regard to emulation is that the simulation model and process

control system with their different clocks for the scheduled arrival times of events have to be synchro-

nized. Thereto, a variation of the blocking rendezvous pattern (Douglass 2004) was implemented, ensur-

ing correct chronologies in the transmission of the orders and at the same time allowing execution of a

simulation model at least a magnitude faster than real time. Thus, the behavior in the simulation model is

independent of simulation speed. This is further supported by not simply employing the OPC protocol

(Open Platform Communications) for the communication link between process control and control layers,

but instead an own implementation.

Our emulation approach does not contain any stochastic processes nor are uncertainties modeled in

another way. The reason is that on the one hand the material flow in the model is controlled by the plant

server. This means that orders for feeding the plant with raw panels or for processing the parts are gener-

ated by the plant server. Values for different parameters like, for example, cutting speed or transportation

speed of a roller track are either included in the orders from the plant server or fixed and stored in the sys-

tem model and hence available in the virtual control and field layer. On the other hand, the performance

indicators are calculated without taking into account probabilities of system errors, as required by the

manufacturer of cut-to-size plants who uses the here presented emulation tool.

5 PLANT LAYOUTS WITH FEEDBACK

In this section, an exemplary plant layout with feedback is presented and discussed. Such plant layouts re-

sult from the corresponding cut-to-size plants being used for make-to-order production: Processing small

quantities, with the special case of lot-size one, results in the cutting patterns becoming more complex

(see Figure 4). Cutting-pattern complexity is formalized and quantified in subsection 6.1 below (see Fig-

ure 9 for illustration).

Waste

Finished part

Figure 4: Panel cutting patterns (simple on the left vs. more complex on the right)

While cut-to-size plants featuring a linear layout, with two saws of orthogonal placement – one for

rip-cuts and one for cross-cuts, respectively (see Figure 1), can process panels according to cutting pat-

terns such as depicted in Figure 4 on the left, the manufacturing of panels after more complex cutting pat-

terns, like the one shown on the right, exceeds the capabilities of such plants. To overcome this limitation,

plant layouts which feature material flows with feedback have been introduced.

In the context of cut-to-size plants, a feedback in the material flow means that panels which have been

cut by a certain saw will be allocated to the same saw again for further partitioning. Referring to Figure 5,

for example, one or more of the panels which result from rip-cutting a raw panel carried out by saw 1 can

be assigned to a feedback conveyor, in order to have their cross cuts performed by saw 1, too. More gen-

erally speaking, a panel (or parts of it) can be fed back to the same saw again and again as often as re-

quired to fulfill the specification given by a certain complex cutting pattern.

The downside of feedbacks in the material flow (saw circles SC1, SC2, and SC3 in Figure 5) towards

realizing more complex cutting patterns is that they may cause deadlocks (Coffman, Elphick, and Sho-

shani 1971). Hence, cases may occur such that there is a set S of panels, where each panel is occupying a

plant unit and waiting for the release of another unit for its advancement. However, at the same time, the

requested units are held by other parts of the set S and thus cannot be freed.

2569


SC3 SC1SC2

S1S3 S2

F

L

D2

D5

D1

D4D3

a

b

c

e

d

fji

gh

k

m

l

Material flow

Roller track

Automaton

Saw

S … Saw

F … Feeding

L … Leaving the plant

D … Decision

SC … Saw circle

Figure 5: Plant layout with feedbacks

Beyond feedbacks in the material flow, also the flexible routing of parts (De Toni and Tonchia 1998)

has newly been established in the control of the material flow for cut-to-size plants. Although there are

constraints concerning the choice of the saws due to technical specifications (for example, the maximal

cutting lengths of saw 2 and 3 in Figure 5 are only 2.3 m respectively, whereas the one of the first saw is

4.3 m), the decision about which saw is to be used next is made in real time on the basis of the current

state of the plant in general. Consequently, the way of a part through the plant is not yet determined at the

time when the part enters the plant. As a result, the material flow is not predictable anymore, as opposed

to cut-to-size plants with simpler (i.e. linear) layouts.

The material flow illustrated in detail in Figure 5 is reduced to its essential stages and presented in

Figure 6.

Leaving the plant

Feeding

Saw 2

Saw 3

Saw 1

Figure 6: Abstracted material flow of the plant shown in Figure 5

Compared to Figure 1, sorting and stacking areas are not incorporated into the plant layout, because

these functions are not as crucial in regard to production of smaller batch sizes as they are for mass pro-

duction.

The other two issues to be considered with regard to cut-to-size plants with feedbacks in the material

flow admittedly also concern cut-to-size plants featuring linear layouts. However, they have to be particu-

larly taken into account with respect to the possible occurrence of deadlocks in cut-to-size plants. The

2570


first of these features is the divergent structure of the product (Günther and Tempelmeier 2012). As the

term ‘cut-to-size plant’ already indicates, therewith denoted manufacturing systems serve to divide up raw

panels. More precisely, two smaller panels arise out of one panel during the process of cutting. Thus, the

number of panels in the system can increase even though there are no new raw panels fed into the plant.

The second characteristic of cut-to-size plants to be newly considered concerns the capacity of a plant

unit. A unit, the purpose of which is to transport panels through the plant, usually consists of one roller

track (see legend of Figure 5). Depending on the size of the panels, a varying number of panels fit on a

roller track and hence the capacity of the plant unit is variable.

For example, if there aren’t any restrictions regarding the quantity of panels on a roller track, panels

can be allocated to the unit as long as there is enough space. Vice versa, if the number of panels located

on a unit must not exceed one, the situation illustrated in Figure 7 can occur: One panel (p1) is longer than

the roller track (r3) to which it is assigned, and therefore still partially lies on the preceding unit (r2). That

is why the next part (p2) cannot enter roller track r2 even though there is no panel assigned to this unit.

Thus, the current capacity of plant unit r2 is zero.

r1 r2 r3

p1

p2

Figure 7: Three roller tracks and two panels

Multiple roller tracks can be consecutively combined for the purpose of locally buffering panels, as it

is the case for the cut-to-size plant presented in Figure 5. For each roller track the maximal capacity is one

(due to technical peculiarities and procedural specifications) and, mostly, they are of short length.

6 TEST-DRIVEN DESIGN OF ZERO-DEADLOCK PLANTS

As compared to cut-to-size plants with linear layouts (see Figure 1), where the ‘circular wait’ condition of

the four necessary conditions for a deadlock to occur (Coffman, Elphick, and Shoshani 1971) never holds,

plants with layouts such as discussed in the preceding section (see Figures 5 and 6) possibly run into

deadlocks. For dealing with these deadlocks, the deadlock avoidance strategy (Li, Wu, and Zhou 2012)

has been chosen for implementation in the plant server software. With this strategy, decisions are made in

real time and are based on the analysis of the current state of the system, i.e. the plant. The strategy is in-

tended to not only avoid that deadlocks may occur but also to contribute to that a high utilization of the

saws is achieved. According to Zajac (2004), deadlock avoidance is the preferable strategy for deadlock

handling in automated manufacturing systems, as opposed to the deadlock-prevention and detection-and-

recovery strategies: Deadlock prevention is unfavorable, because it often results in poor utilization of

manufacturing units due to extensive restrictions on system operation. Deadlock detection and recovery is

not applicable in this case, because preemption of units from a number of deadlocked processes is not al-

lowed while running a cut-to-size plant (Zajac 2004).

In what follows, the deadlock avoidance strategy for plant layouts with feedbacks in the material flow

and its testing by means of emulation are presented in more detail.

6.1 Deadlock-Avoidance Strategy: Implemented Decision Criteria

In order to achieve the objective of operating a cut-to-size plant without deadlocks and at the same time

with a high utilization of saws, a deadlock-avoidance strategy is implemented within the process control

layer. As just described, it analyzes the current state of the plant and, on this basis, comes to a decision.

There are two different types of decisions. Firstly, after the release of an automaton which feeds a saw cir-

cle, the question has to be answered whether a new panel shall be fed into the saw circle or whether a

2571


panel already located in the saw circle shall be allocated to the automaton (decisions D1, D2, and D3 in

Figure 5). To this end, the discrete degree of filling of the saw circle is used as decision criterion; its cal-

culation is described below. The second decision type concerns the assignment of panels to particular

saws in real time – in short, the above mentioned flexible routing of parts. The decision between saw 1

and the other two saw circles is made before a not yet finished panel leaves the automaton placed after

saw 1 in material flow direction (decision D4 in Figure 5). In case the part will not be allocated to saw 1

again, the next decision is between saw 2 and 3 (decision D5 in Figure 5). These decisions depend on the

number of pending processing steps of all the panels currently located within a saw circle. They mainly

affect the utilization of the saws but they are also relevant with regard to the avoidance of deadlocks, be-

cause of the mentioned constraints in relation to the choice of the saws (section 5). Figure 8 shows the ab-

stracted material flow expanded with the discussed decisions.

F S1

LS3

S2D1 D5

D3

D4 D2

Figure 8: Abstracted material flow with decisions

Both types of decision rules will now be described in more detail. For the decision, whether a new

part shall be fed into a saw circle, the discrete degree of filling of this saw circle is calculated. That is

achieved by way of the following four steps:

Listing of the units of the saw circle: The plant units are listed in the reverse order as to the direc-

tion of the material flow, starting with the roller track from which the panels already located in

the saw circle are transferred to the automaton, which also feeds the saw circle. For example, the

sequence of the units of saw circle 1 in Figure 5 arises from the material flow arrows indicated by

l, f, c, and m, from the arrowhead to the end of the shaft respectively.

Listing of the panels of the saw circle: The panels currently located in the saw circle are regis-

tered in the same order as the units. However, in order to take into account the divergent structure

of the product, a panel is actually not listed itself. Instead the parts which will emerge from the

next passage of the panel through a saw are listed.

Determination of the sizes of the listed panels: In this context the size of a panel is defined as the

length of the longer of both lateral dimensions, regardless of how the panel will be situated on a

unit: the longer side parallel to or orthogonal to the material flow direction. The size of a plant

unit is given by its length in material flow direction.

Assignment of the listed panels to the units: Both, the list of plant units and the list of panels are

related to each other by successively assigning a panel to the next unit, the capacity of which is

not exhausted. The panel size determined in the preceding step is compared with the size of the

unit in order to determine the capacity of the next units in the list. This is because the panel might

lie on them (see the discussion on ‘variable capacity’ in section 5).

The discrete degree of filling is defined as the number of the listed units, the capacities of which are

exhausted and it serves as decision criterion. If this number is greater than a pre-assigned threshold value,

no new panel is fed into the saw circle.

On the other hand, the decision rule for the choice of saw compares the number of the pending pro-

cessing steps of the saw circles and prioritizes the saw circle with the fewest processing steps: For each

panel p out of the set P of all panels, which are located in a saw circle, the number of the pending pro-

2572


cessing steps is determined, and then, they are added up. The pending processing steps of one panel p

consist of the number of cuts Cp still to be done plus the number of the corresponding saw passages SPp.

One saw passage comprises all the cuts of a panel, which can be executed without the need to turn this

panel. Thus, the formula for the calculation of the number of the pending processing steps of a saw circle

(PPS) reads

.

Pp

pp SPCPPS

For illustration purposes, the number of pending processing steps of the panel, the cutting pattern of

which is shown in Figure 9, will now be computed: The number of cuts is sixteen (each cut is numbered

in Figure 9), and five saw passages are necessary for the partitioning of the panel. The cuts with numbers

1, 2, and 3 take place when the panel passes a saw for the first time. The two small parts at the edges are

waste material and are disposed immediately after sawing, whereas each of the other two parts passes a

saw again, but rotated by ninety degrees. They are saw passages number 2 (cuts 4 to 7) and 3 (cuts 10 to

14). The cuts with numbers 8 and 9 belong to the fourth saw passage and the ones with numbers 15 and

16 to the fifth one. Therefore, the number of pending processing steps of the depicted panel is 16 + 5 =

21.

1 2 3

4

5

7

16

6

15

14

13

1211

10

98

Figure 9: Another complex panel cutting pattern

6.2 Balancing Plant Performance and Absence of Deadlocks

As said above, the main intention of deploying the emulation tool is to enable ‘cut-to-size plant’ domain

experts to define a high utilization of deadlock-free saw circles, for then evaluating them against prede-

fined critical performance indicators (such as cycle times, throughput rates, and utilization of plant units)

in order to identify an optimal solution. Here, it should be noted that simulation is an evaluation technique

and does not automatically produce an optimal solution, unless simulation runs are controlled by an ex-

ternal search loop (see below).

In the process control layer the decision algorithm for deadlock avoidance is activated for the relevant

plant units, each of which also contains decision rule descriptions about degree of filling and/or pending

processing steps.

After starting the emulation, orders will be served through the process control layer. Before the pro-

cess control layer sends new orders to designated plant units of the virtual control layer, all plant units

with the activated decision algorithm compute the current degree of filling and/or pending processing

steps. The computed results of the decision algorithm control and rearrange the new orders for registering

all activities as tasks in the virtual control layer. The emulation runs and the subsequent analysis of results

can serve as an input for modifications of parameter values of the decision rules. The following procedure

illustrates the iterative approximation to an optimal solution of a plant model:

1. Initialization: Set low values for decision rules so that a deadlock-free saw circle with low utiliza-

tion will be achieved.

2. Always run an emulation job with constant cutting patterns and device settings for mechanical in-

formation and velocities and record data for the new scenario.

2573


3. If a deadlock-free emulation job is available, then analyze the utilization of plant units and in-

crease the values for the decision rules. Go to step 2 and run the emulation again.

4. If a deadlock has occurred during an emulation job, then decrease the values for the decision

rules. Go to step 2 and generate a new scenario.

5. If a deadlock has occurred during an emulation job and a valid scenario of lower value settings

for decision rules is already available, the optimal solution of the plant model has been found.

6. If a deadlock has occurred during an emulation job and decreasing the values for decision rules is

not possible, then the capacity of the unit which is used as buffer of the saw circles has to be ad-

justed for a new plant model.

The optimal solution found is based on the assumption that for real product manufacturing, there are

valid plant-specific cutting patterns and hence predefined constraints, such as maximal size of panel, size

of part of panel or number of cuts per cycle, are used for pattern formation for parts lists.

7 RESULTS AND FINDINGS

At the moment, the above described, implemented, and deployed emulation tool provides relevant infor-

mation, models, and methods for acquisition, design and planning, as well as testing and implementation

stages of cut-to-size plants, where the underlying models are consistently used across process stages. The

demonstrated techniques facilitate the anticipation of real-life plant behavior by means of a virtual sys-

tem. This in turn allows plant experts to model, emulate, and animate sequences of cut-to-size plant oper-

ations, without the need for in-depth simulation expertise: System models are built following a modeling

process, characterized by plant-unit template configuration und drag-and-drop style system construction.

An automatically generated simulation model allows running emulation jobs, where results are logged

and enable a plant expert to meaningfully analyze critical performance indicators (cycle times, through-

put, and utilization).

First of all, the presented emulation tool is successfully employed in the rather complex sales proce-

dures typical in regard to cut-to-size plants, with a proven track record of enabling a key competitive ad-

vantage via the cost-effective, reliable, and accurate calculation of all the relevant performance indicators,

potential clients are interested in. The role of emulation in this process has been extensively discussed in

our previous work. Second, and of no less importance, the complexities of cut-to-size plants with feed-

backs in the material flow are kept manageable with the presented emulation tool, processes, and proce-

dures. This is another key advantage not only concerning sales procedures but also as to providing a vir-

tual, high-level plant-control system test bed. Summarizing, technical expertise, constancy, and reliability

can be demonstrated through virtual plants.

The emulation tool supports the plant experts in planning processes, in modeling and evaluating cut-

ting areas with feedback and in validation of deadlock-free solutions. The comparison of several emula-

tion scenarios allows the plant experts to optimize the solution according to the procedure explained in

section 6.2. Overall, with the presented approach, tools, models, and techniques costs and time can be

saved in the planning process. The planning instrument not only enables plant experts to model layouts

with feedback with multiple parallel cutting areas but during implementation of the real plant, the instru-

ment serves an incremental implementation of the cutting areas. For example, in the ramp-up of a further

cutting area, the tool enables the plant experts to find an optimal solution for the plant extension regarding

the absence of deadlocks. Also, the tool can serve as training instrument to get familiar with the control

system and plant processes. While operating the plant, it is possible to perform impact analysis of modifi-

cations of the virtual as well as the real system. This allows for an efficient analysis and resolution of er-

rors, without disrupting the normal course of business. With the realization of the described emulation

tool, a planning instrument becomes feasible which reduces acquisition and start-up times for cut-to-size

plants and continuously tests and optimizes the control processes.

More activities to further developing and extending the emulation tool are planned in the course of

this year, amongst them not only targeting the upper-level MES software but also the lower-level soft-

2574


ware, i.e. PLC code. Currently, real plant units can often only be tested separately and not as a combined

plant system. The idea is to create a complex system where a combination of reality and simulation can be

achieved. Therewith, real units of the plant could be tested within the overall virtual system. This ap-

proach will help the plant experts to be able to find failures on plant units or in the material flow much

earlier as without the emulation tool.

ACKNOWLEDGMENTS

This paper discusses the results and findings of a research project within the K-Project ‘Integrated Deci-

sion Support Systems for Industrial Processes (ProDSS)’ which has been financed under the Austrian

funding scheme COMET (COMpetence centers for Excellent Technologies).

REFERENCES

Coffman, E. G., M. J. Elphick, and A. Shoshani. 1971. “System Deadlocks.” Computing Surveys 3:67–78.

De Toni, A., and S. Tonchia. 1998. “Manufacturing Flexibility: A Literature Review.” International

Journal of Production Research 36:1587–1617.

Dobler, M., M. Saler, and L. Maerz. 2008. “Distribution of Concurrent Simulation Runs on a Service-

oriented Network Structure.” In Advances in Simulation for Production and Logistics Applications,

Edited by M. Rabe, 509–517. Stuttgart: Fraunhofer IRB Verlag.

Douglass, B. P. 2004. Real Time UML: Advances in the UML for Real-time Systems. Boston: Addison-

Wesley.

Günther, H.-O., and H. Tempelmeier. 2012. Produktion und Logistik. Berlin, Heidelberg: Springer-

Verlag.

Günthner, W., and M. ten Hompel. 2010. Internet der Dinge in der Intralogistik. Berlin, Heidelberg:

Springer-Verlag.

Hillbrand, C., and S. Schmid. 2011. “Simulation of Multimodal Logistics Networks.” In Proceedings 25th

European Conference on Modelling and Simulation, Edited by T. Burczynski, J. Kolodziej, A. Byr-

ski, and M. Carvalho, 594–600.

Li, Z., N. Wu, and M. Zhou. 2012. “Deadlock Control of Automated Manufacturing Systems Based on

Petri Nets–A Literature Review.” IEEE Transactions on Systems, Man, and Cybernetics–Part C: Ap-

plications and Reviews 42:437–462.

Maerz, L., and M. Saler. 2008. “An Analysis, Planning and Decision Tool for Warehouse Applications.”

In Advances in Simulation for Production and Logistics Applications, Edited by M. Rabe, 237–245.

Stuttgart: Fraunhofer IRB Verlag.

McGregor, I. 2002. “The Relationship Between Simulation and Emulation.” In Proceedings of the 2002

Winter Simulation Conference, Edited by E. Yücesan, C.-H. Chen, J. L. Snowdon, and J. M. Charnes,

1683–1688. Piscataway, New Jersey: Institute of Electrical and Electronics Engineers, Inc.

Mertens, P. 2001. Lexikon der Wirtschaftsinformatik. Berlin, Heidelberg: Springer-Verlag.

Meyer, T., C. Pöge, and G. Mayer. 2012. “Integration of Emulation Functionality into an Established

Simulation Object Library.”In Proceedings of the 2012 Winter Simulation Conference, Edited by C.

Laroque, J. Himmelspach, R. Pasupathy, O. Rose, and A. M. Uhrmacher, 2864–2874. Piscataway,

New Jersey: Institute of Electrical and Electronics Engineers, Inc.

Okolnishnikov, V. 2011. “Development of process control systems with the use of emulation models.” In-

ternational Journal of Mathematics and Computers in Simulation 5:553–560.

Schoech, R., S. Schmid, and C. Hillbrand. 2011. “Simulation Based Decision Support System for Cut-to-

size Plants.” In Proceedings 25th European Conference on Modelling and Simulation, Edited by T.

Burczynski, J. Kolodziej, A. Byrski, and M. Carvalho, 394–400.

Schoech, R., S. Schmid, and C. Hillbrand. To Appear. "Optimising Plant Layout Decisions Based on

Emulation Models - Technical Framework and Practical Insights." International Journal of Simula-

tion and Process Modelling, Special Issue on: "Research Challenges in Developing Advanced Solu-

2575


tions in Industry and Supply Chains Modelling and Simulation for Building 21st Century Enterpris-

es." http://www.inderscience.com/info/ingeneral/forthcoming.php?jcode=ijspm.

Seidel, S., U. Donath, and J. Haufe. 2012. “Towards an Integrated Simulation and Virtual Commissioning

Environment for Controls of Material Handling Systems.” In Proceedings of the 2012 Winter Simula-

tion Conference, Edited by C. Laroque, J. Himmelspach, R. Pasupathy, O. Rose, and A. M. Uhr-

macher, 2852–2863. Piscataway, New Jersey: Institute of Electrical and Electronics Engineers, Inc.

Zajac, J. 2004. “A Deadlock Handling Method for Automated Manufacturing Systems.” Annals of the

CIRP 53:367–370.

Zeigler, B. P., and H. S. Sarjoughian. 2003. “Introduction to DEVS Modeling & Simulation with JAVA:

Developing Component-based Simulation Models.” http://www.cs.gsu.edu/xhu/CSC8840/DEVS

JAVA_Manuscript.pdf.

AUTHOR BIOGRAPHIES

RUTH FLEISCH is research associate at V-Research. After receiving her diploma in mathematics from the Uni-

versity of Innsbruck she joined V-Research in 2010. Her research focuses on discrete-event systems as well as RFID

technology. Her email address is [email protected].

ROBERT SCHOECH is project manager at V-Research, an Austrian competence center for industrial research and

development. Within the research area “Technical Logistics” the research center focuses on methods to support

complex decisions in manufacturing and logistics processes. Robert Schoech’s work focuses on simulation-based

systems for industrial processes. His research interests also include track & trace solutions based on GPS, GSM, and

RFID technology. His email address is [email protected].

THORSTEN PRANTE is a project manager at V-Research and the director of the research program “ProDSS” (In-

tegrated Decision Support Systems for Industrial Processes). He has worked on man-machine interaction in the con-

text of ubiquitous computing for a number of years and is a co-founder of the open-source project Zeitgeist on activ-

ity-based information management. Thorsten’s current work focusses around balancing interaction and automation

in the field of engineering design informatics. His email address is [email protected].

ROBERT PFLEGERL is software developer at Schelling, a leading manufacturer of cut-to-size plants. His work

focuses on the preparation and administration of process data for the automated procedures of the plants, the devel-

opment of graphical user interfaces for the visualization of the states of the plant and diagnostic data, as well as the

capture and analysis of manufacturing data. His email address is [email protected].

2576

consistent use of emulation across different stages …

Documents